Variant Reclassification Analysis

Investigation of Pathogenic to Benign Reclassifications - Expert Curated Dataset

Generated: 2025-10-21 03:43:07

⚠️ Dataset Note: This analysis correctly separates validated P→B reclassifications (exclude=false, n=22) from problematic ones identified by expert curation (exclude=true, n=0).

Table of Contents

Summary Statistics

Total Variants in Dataset
35
Validated P→B Variants
22
Excluded P→B Variants
0
Total P→B Candidates
22
Genes with P→B
10
Median Time to Benign Reclassification
80.4 mo
Inter-lab Flips
100.0%
Top Reason
Population frequency

Exclusion Analysis

Analysis of excluded P→B variants and special notes for included variants.

Included vs Excluded P→B Variants
Distribution of P→B candidates by validation status.

Gene & Variant Landscape

Distribution of validated P→B reclassifications across genes and their characteristics.

Top Genes by P→B Variant Count
Genes with the highest number of validated variants reclassified from pathogenic to benign.

Classification Flow & Timing

Temporal patterns in validated variant reclassifications.

Reclassification Timeline
Temporal distribution of initial P/LP and subsequent B/LB classifications for validated variants.

Reasons for Reclassification

Inferred reasons driving validated P→B reclassifications based on ACMG evidence codes.

Primary Reasons for Reclassification
Distribution of reason categories across validated P→B variants.
Reason Distribution by Submitter
Heatmap showing which submitters contribute to each reclassification reason (validated variants only).

Submitter Dynamics

Analysis of submitter patterns in validated reclassifications.

Top Submitters on Benign Side
Laboratories most frequently submitting benign classifications for validated P→B variants.

Finding #1: Time to Correction – Median 6 Years

The time between initial pathogenic classification and benign reclassification spans years, with a median of approximately 6 years. This lag represents a critical window where patients may receive inaccurate genetic counseling.

Summary Statistics:

Time to Correction Distribution
Time to Correction by Gene
Fastest Corrections (< 24 months)
  • COL4A5: NM_033380.3(COL4A5):c.2692A>G (p.Met898Val) — 4.0 months (Yale Center for Mendelian Genomics, Yale University → PreventionGenetics, part of Exact Sciences)
  • TSC2: NM_000548.5(TSC2):c.1081C>G (p.Leu361Val) — 19.6 months (Liping Wei Laboratory, Peking University → GeneDx)
Slowest Corrections (> 120 months)
  • PKD1: NM_001009944.3(PKD1):c.971G>T (p.Arg324Leu) — 237.6 months (OMIM → ARUP Laboratories, Molecular Genetics and Genomics, ARUP Laboratories)
  • PKD1: NM_001009944.3(PKD1):c.971G>T (p.Arg324Leu) — 237.6 months (OMIM → ARUP Laboratories, Molecular Genetics and Genomics, ARUP Laboratories)
  • HNF1A: NM_000545.8(HNF1A):c.1748G>A (p.Arg583Gln) — 168.9 months (OMIM → Ambry Genetics)
  • NPHS2: NM_014625.4(NPHS2):c.59C>T (p.Pro20Leu) — 154.1 months (OMIM → Women's Health and Genetics/Laboratory Corporation of America, LabCorp)
  • NPHS2: NM_014625.4(NPHS2):c.59C>T (p.Pro20Leu) — 154.1 months (OMIM → Women's Health and Genetics/Laboratory Corporation of America, LabCorp)
Clinical Impact
  • Average patient under false-positive diagnosis: 6–8 years
  • Cascade testing consequences: family members screened/managed on a false premise
  • Insurance/employment discrimination persists

Finding #2: BS1/BA1 – The Game-Changer

Population frequency evidence (BA1/BS1) is the dominant driver of pathogenic-to-benign reclassifications. As large-scale population databases like gnomAD have grown, variants once thought rare enough to be pathogenic are now recognized as benign polymorphisms.

BA1/BS1 prevalence increased dramatically from 0.0% on the P/LP side to 50.0% on the B/LB side.

Fisher's Exact Test: Odds Ratio = inf, p-value = 1.84e-04

BA1/BS1 Usage: Before vs After
The Problem with PM2
"Absent from controls" ≠ pathogenic. Pre-2016 datasets (e.g., ExAC) had limited ancestry. Post-2020 gnomAD includes large, diverse cohorts; many variants formerly "absent" now occur at 1:200–1:1000 in healthy populations. Population evidence (BS1/BA1) supersedes legacy PM2 calls.

Finding #3: From Generic ACMG → Gene-Specific Specifications

The evolution from the 2015 ACMG/AMP guidelines to gene-specific VCEP (Variant Curation Expert Panel) specifications has fundamentally changed how evidence is weighted and interpreted.

Approximately 22.7% of reclassifications involve both a change in guideline and a shift in evidence bucket composition, highlighting the impact of gene-specific VCEP specifications.

Timeline of Guideline Publications
All unique guidelines extracted from JSON with dates parsed from guideline names or earliest usage dates.
ACMG 2015 vs Gene-Specific Examples
Framework Key Changes Example Genes
ACMG 2015 (generic) One-size-fits-all evidence codes All genes
BRCA1/2 ENIGMA (2020) Adjusted functional assay thresholds, refined splicing rules BRCA1, BRCA2
TSC1/2 Zhou (2019) LOF mechanism-specific criteria TSC1, TSC2
Kidney VCEP (in development) Disease-specific population frequency thresholds PKD1, PKD2, COL4A5

Finding #4: When Algorithms Disagree (or Change Their Mind)

Computational prediction tools (PP3 for pathogenic, BP4 for benign) are inherently unstable. As algorithms are retrained on new data or replaced with newer models, the same variant can flip from computationally predicted pathogenic to benign.

Computational evidence shifted in 5 variants with PP3 (P/LP side) and 10 variants with BP4 (B/LB side). Below are case studies where the same variant went from PP3-supported pathogenic to BP4-supported benign:

PP3→BP4 Case Studies (Longest Time to Correction)
Gene Variant Time (mo) P/LP Date P/LP Submitter P/LP Codes B/LB Date B/LB Submitter B/LB Codes
BRCA2 NM_000059.4(BRCA2):c.8851G>T (p.Ala2951S... 61.2 2020-04-08 Center of Medical Genetic PM1,PM2,PP3,PP4 2025-05-16 Ambry Genetics BP2,BP4,BP5,BP7,BS1,BS2,BS3,BS
NPHS1 NM_004646.4(NPHS1):c.559G>A (p.Val187Met... 53.5 2018-09-16 Gharavi Laboratory, Colum PM2,PP3,PP4 2023-03-02 Labcorp Genetics (formerl BA1,BP4,BS1
TSC2 NM_000548.5(TSC2):c.4639G>A (p.Val1547Il... 45.3 2017-04-11 OMIM PM2,PP3,PS3 2021-01-19 Ambry Genetics BP1,BP4,BP5,BS1,BS2,BS3,BS4
PAX2 NM_000278.5(PAX2):c.1058A>C (p.Gln353Pro... nan N/A Precision Medicine Center PM1,PP1,PP3,PP4,PP5 2020-01-22 PreventionGenetics, part BP4,BP7,BS1
Timeline of PP3/BP4 Variants

Finding #5: When Lab Assays Contradict Each Other

Functional evidence is often treated as highly reliable (PS3 for pathogenic functional studies, BS3 for benign). However, the data show that variants with PS3 evidence can later acquire BS3 evidence, suggesting that single functional assays may be insufficient or that assay conditions matter greatly.

Functional evidence reversed in 2 variants with PS3 (P/LP side) and 3 variants with BS3 (B/LB side). Below are case studies where functional assays contradicted each other:

PS3→BS3 Case Studies (Longest Time to Correction)
Gene Variant Time (mo) P/LP Date P/LP Submitter P/LP Codes B/LB Date B/LB Submitter B/LB Codes
TSC2 NM_000548.5(TSC2):c.4639G>A (p.Val1547Il... 45.3 2017-04-11 OMIM PM2,PP3,PS3 2021-01-19 Ambry Genetics BP1,BP4,BP5,BS1,BS2,BS3,BS4
Functional Assay Standards: Then vs Now
Era/Standard Typical Assay Limitations Modern Approach
Legacy PS3 (single-variant) Cell viability, Western blot, localization Small N, variable conditions, publication bias
Current BS3 (validated, multiplex) MAVEs, HDR assays, orthogonal validation Assay validation, controls, reproducibility

Detailed Variant Table

Complete list of validated P→B variants with classification details and inferred reclassification reasons.

Showing 22 validated P→B variants (exclude=false) sorted by gene and time to reclassification. Table is scrollable.

gene variant type first_plp_date first_blb_date first_plp_submitter first_blb_submitter time_to_benign_months interlab plp_codes blb_codes inferred_reasons exclude_reason
BRCA2 NM_000059.4(BRCA2):c.8187G>T (p.Lys2729Asn) missense 2002-07-26 2011-03-14 OMIM Sharing Clinical Reports Project (SCRP) 103.6 True Other/Unknown Special Note: A submission from the Online Mendelian Inheritance in Man (OMIM) for this variant is associated with a disease identifier that was once considered pathogenic. However, this does not mean that OMIM submitted a classification of pathogenic for this specific variant. The reason is a known discrepancy in how OMIM data was historically displayed on ClinVar.
BRCA2 NM_000059.4(BRCA2):c.7879A>T (p.Ile2627Phe) missense 2012-03-11 2019-09-01 Sharing Clinical Reports Project (SCRP) King Laboratory, University of Washington 89.7 True Other/Unknown Special Note: This variants has a complcated history of calssifications. Although the ClinGen Expert Panel classified it as Pathogenic for BRCA2-related cancer predisposition on 2024/06/11, there is a hiden record in vcv xml file saying a new submitter submitted the classification as benign on 2025/09/07 (ClinVar might not have time to review and update it). Also in its publically shown history record, King Laboratory, Univeristy of Washington classified it as Benign on 2019/09/01 with a cited publication.
BRCA2 NM_000059.4(BRCA2):c.891_899delinsGATACTTAG (p.Thr298_Val300delinsIleLeuArg) missense 2017-04-20 2023-03-23 Department of Pathology and Molecular Medicine, Queen's University University of Washington Department of Laboratory Medicine, University of Washington 71.1 True PM1,PM2,PP4 BP1 Guideline/model change N/A
BRCA2 NM_000059.4(BRCA2):c.8851G>T (p.Ala2951Ser) missense 2020-04-08 2025-05-16 Center of Medical Genetics and Primary Health Care Ambry Genetics 61.2 True PM1,PM2,PP3,PP4 BP2,BP4,BP5,BP7,BS1,BS2,BS3,BS4 Population frequency, New benign functional evidence, Computational shift N/A
BRCA2 NM_000059.4(BRCA2):c.2623G>C (p.Val875Leu) missense 2020-04-08 2023-03-23 Center of Medical Genetics and Primary Health Care University of Washington Department of Laboratory Medicine, University of Washington 35.4 True BP4,PM1,PM2,PP1,PS1 BP4,BS1 Population frequency, Guideline/model change N/A
COL4A5 NM_033380.3(COL4A5):c.2692A>G (p.Met898Val) missense 2019-02-14 2019-06-15 Yale Center for Mendelian Genomics, Yale University PreventionGenetics, part of Exact Sciences 4.0 True Other/Unknown N/A
HNF1A NM_000545.8(HNF1A):c.1748G>A (p.Arg583Gln) missense 2002-10-01 2016-10-27 OMIM Ambry Genetics 168.9 True BP1,BP4,BP5,BS1,BS2,BS3,BS4 Population frequency, New benign functional evidence N/A
HNF1A NM_000545.8(HNF1A):c.92G>A (p.Gly31Asp) missense 2005-03-01 2017-03-01 OMIM GeneDx 144.0 True BP4,BS1 Population frequency N/A
NPHS1 NM_004646.4(NPHS1):c.559G>A (p.Val187Met) Missense variant in extracellular domain of nephrin protein 2018-09-16 2023-03-02 Gharavi Laboratory, Columbia University Labcorp Genetics (formerly Invitae), Labcorp 53.5 True PP4 Other/Unknown N/A
NPHS1 NM_004646.4(NPHS1):c.559G>A (p.Val187Met) missense 2018-09-16 2023-03-02 Gharavi Laboratory, Columbia University Labcorp Genetics (formerly Invitae), Labcorp 53.5 True PM2,PP3,PP4 BA1,BP4,BS1 Population frequency (BA1 threshold), Computational shift, Guideline/model change N/A
NPHS1 NM_004646.4(NPHS1):c.559G>A (p.Val187Met) missense 2018-09-16 2023-03-02 Gharavi Laboratory, Columbia University Labcorp Genetics (formerly Invitae), Labcorp 53.5 True Other/Unknown N/A
NPHS2 NM_014625.4(NPHS2):c.59C>T (p.Pro20Leu) missense 2003-05-01 2016-03-04 OMIM Women's Health and Genetics/Laboratory Corporation of America, LabCorp 154.1 True PM1,PM2,PP1,PS4 BP4,BP6,BS1,BS2 Population frequency Special Note: variant was initially considered pathogenic. This early classification, however, has since been re-evaluated and corrected based on further research and genetic analysis.
NPHS2 NM_014625.4(NPHS2):c.59C>T (p.Pro20Leu) missense 2003-05-01 2016-03-04 OMIM Women's Health and Genetics/Laboratory Corporation of America, LabCorp 154.1 True BS1,BS2 Population frequency Special Note: Initial association with disease (2000s): Early studies, including one in 2000, identified this variant in patients with steroid-resistant nephrotic syndrome.
PAX2 NM_000278.5(PAX2):c.1058A>C (p.Gln353Pro) missense 2020-01-22 Precision Medicine Center, Zhengzhou University PreventionGenetics, part of Exact Sciences NaN True PM1,PP1,PP3,PP4,PP5 BP4,BP7,BS1 Population frequency, Computational shift, Guideline/model change Special Note: The Likely pathogenic submission is a flagged submission, with reason of claim with insufficient supporting evidence. From ClinVar, it says Explanation for multiple conditions: Uncertain. The variant was classified for several related diseases, possibly a spectrum of disease; the variant may be associated with one or more the diseases.
PAX2 NM_000278.5(PAX2):c.1058A>C (p.Gln353Pro) missense 2020-01-22 Precision Medicine Center, Zhengzhou University PreventionGenetics, part of Exact Sciences NaN True PM1,PP1,PP3,PP4,PP5 Other/Unknown N/A
PKD1 NM_001009944.3(PKD1):c.971G>T (p.Arg324Leu) missense 1999-07-01 2019-04-19 OMIM ARUP Laboratories, Molecular Genetics and Genomics, ARUP Laboratories 237.6 True Other/Unknown N/A
PKD1 NM_001009944.3(PKD1):c.971G>T (p.Arg324Leu) missense 1999-07-01 2019-04-19 OMIM ARUP Laboratories, Molecular Genetics and Genomics, ARUP Laboratories 237.6 True Other/Unknown N/A
RET NM_020975.6(RET):c.2372A>T (p.Tyr791Phe) missense 2006-08-28 2012-07-13 Clinical Molecular Genetics Laboratory, Johns Hopkins All Children's Hospital Biesecker Lab/Clinical Genomics Section, National Institutes of Health 70.5 True Other/Unknown N/A
SCNN1G NM_001039.4(SCNN1G):c.589G>A (p.Glu197Lys) missense 2008-05-28 2017-04-27 OMIM Illumina Laboratory Services, Illumina 107.0 True PP1,PS3 BP4,BS1 Population frequency N/A
SCNN1G NM_001039.4(SCNN1G):c.589G>A (p.Glu197Lys) missense 2008-05-28 2017-04-27 OMIM Illumina Laboratory Services, Illumina 107.0 True BS1 Population frequency N/A
TSC2 NM_000548.5(TSC2):c.4639G>A (p.Val1547Ile) missense 2017-04-11 2021-01-19 OMIM Ambry Genetics 45.3 True PM2,PP3,PS3 BP1,BP4,BP5,BS1,BS2,BS3,BS4 Population frequency, Functional (benign evidence), Computational shift N/A
TSC2 NM_000548.5(TSC2):c.1081C>G (p.Leu361Val) missense 2018-08-01 2020-03-20 Liping Wei Laboratory, Peking University GeneDx 19.6 True BP4,BS4 Guideline/model change N/A

Methods & Assumptions